Sub-phenotypes of AKI: Do We Have Progress for Personalizing Care?

Abstract

Acute kidney injury (AKI) is the most common form of organ dysfunction occurring in patients admitted to the intensive care unit (ICU) and contributes significantly to poor long-term outcomes. Despite this public health impact, no effective pharmacotherapy exists for AKI. One reason may be that heterogeneity is present within AKI as currently defined, thereby concealing unique pathophysiologic processes specific to certain AKI populations. Supporting this notion, we and others have shown that diversity within the AKI clinical syndrome exists and the ‘one size fits all’ approach by current diagnostic guidelines may not be ideal. A ‘precision medicine’ approach that exploits an individual’s genetic, biologic and clinical characteristics to identify AKI sub-phenotypes may overcome such limitations. Identification of AKI sub-phenotypes may address a critical unmet clinical need in AKI by 1) improving risk prognostication, 2) identifying novel pathophysiology and 3) informing a patient’s likelihood of responding to current therapeutics or establishing new therapeutic targets to prevent and treat AKI. This review discusses the current state of phenotyping AKI and future directions.

Introduction

Acute Kidney Injury (AKI) is common in hospitalized populations and associated with morbidity and mortality. Multiple observational studies have shown a link between AKI and the development of chronic kidney disease, ESRD, cardiovascular disease, and mortality, even in subjects with reported renal recovery¹. Numerous clinical trials have been conducted in an effort to identify disease-modifying therapies, yet no safe and effective pharmacotherapeutics have been identified. One reason for the lack of therapies may be that lumping patients with different AKI risk factors and pathophysiology may hide a treatment signal specific to certain biologically distinct AKI populations.

The Kidney Disease Improving Global Outcomes (KDIGO) consensus group defines AKI as an increase in serum creatinine (SCr) of ≥ 0.3 mg/dl or > 50% of the baseline within a 48-hour period or a 7-day time period. Although this definition has resulted in improved recognition and the discovery of critical concepts applicable to populations of patients, it fails to stratify subjects with different underlying biology, recovery patterns and subsequent outcomes. Identifying meaningful groups within the diverse syndrome of AKI may provide a deeper understanding of disease pathophysiology and aid in discovery of potential therapeutic targets.

Clinical AKI Sub-phenotypes

Despite the broad definition for AKI, clinicians have intuitively “phenotyped” AKI into groups for years: pre-renal AKI (i.e., “volume-responsive”) and intrinsic AKI (i.e., acute tubular necrosis, ATN). Although conceptually straightforward, there is difficulty in reliably identifying these groups. A recent intervention, known as the “Furosemide Stress Test”, seeks to help clinicians identify these two AKI sub-groups. The furosemide stress test challenges renal tubular function by administering high dose furosemide and assessing the subsequent urine output response within two hours of administration, thereby providing an indirect measure of functional nephrons². In addition, a series of pre-clinical and clinical experiments attempted to demonstrate whether pre-renal and ATN sub-groups share similar pathogenesis or are distinct entities³. By developing murine models for pre-renal and intrinsic AKI and performing complete RNA sequencing of mouse kidney tissue, Xu and colleagues identified thousands of genes that were differentially regulated. The investigators then measured 40 corresponding candidate urinary proteins in clinical AKI and found different urinary protein concentrations between patients with volume responsive and intrinsic AKI, suggesting that these two entities were distinct groups with differing pathophysiology.

Several research groups have sought to identify AKI sub-groups based on kidney functional recovery after injury^4,5. In 2016, Bhatraju and colleagues identified two AKI subphenotypes based on the trajectory of serum creatinine, “resolving” and “non-resolving” AKI⁴. Patients with non-resolving AKI had higher hospital mortality and in a subsequent analysis patients with non-resolving AKI had poorer long-term outcomes, such as development and progression of chronic kidney disease, dialysis and death, than patients with resolving AKI⁶. These associations were independent of the traditional KDIGO severity stage of AKI.

Biomarker Based AKI Sub-phenotypes

A number of blood and urinary biomarkers have been shown to predict the development of AKI, such as kidney injury molecule 1, neutrophil gelatinase associated lipocalin, tissue inhibitor of metalloproteinase-2 and insulin-like growth factor binding protein 7, and others⁷. More recently studies have used these biomarkers to describe an emerging AKI phenotype called sub-clinical AKI, in which serum creatinine does not rise but urinary biomarkers are elevated⁸. Alternative approaches to identifying AKI sub-phenotypes include unsupervised clustering analyses of high-dimensional biological data, incorporating multiple biomarkers. In 2018, Bhatraju and colleagues applied latent class analysis methodology to identify two novel AKI sub-phenotypes, AKI-SP1 and AKI-SP2, with the latter associated with greater risk of 7-day renal non-recovery, need for renal replacement therapy and 28-day mortality when compared to AKI-SP1⁹. The investigators then developed a three-biomarker prediction model (ratio of angiopoietin-2/angiopoietin-1 and tumor necrosis factor receptor 1) to identify AKI subphenotypes. Next, the investigators used blood collected prior to randomization to identify AKI sub-phenotypes in the Vasopressin versus Norepinephrine in Septic Shock Study (VASST) clinical trial. Patients with AKI-SP1 had improved mortality with the early addition of vasopressin as opposed to norepinephrine only (27% vs 46%, p=0·02), but no benefit was observed for patients with AKI-SP2 (45% vs 49%, p=0·99) and the p-value for interaction was 0.05 (Figure 1). These findings are notable, as the initial VASST study demonstrated no significant difference in 28 or 90-day mortality rate between vasopressor therapy. Subsequently, Wiersema and colleagues in 2020 applied latent class analysis to the Finnish Acute Kidney Injury Cohort and identified similar AKI sub-phenotypes¹⁰.

Figure 1. Treatment interaction between AKI sub-phenotypes and vasopressor therapy in patients with septic shock.

Panel A shows the patients included in the secondary analysis of the VASST clinical trial. Panel B shows the forest plot of the risk of 90-day mortality. Relative risk estimates were adjusted for APACHE II scores and norepinephrine dose prior to randomization.

Another emerging tool to identify AKI sub-phenotypes includes machine-learning techniques using electronic medical record (EMR) data. Chaudhary and colleagues used deep learning algorithms to extract data from the EMR and then used K-means clustering to identify three sepsis-associated AKI sub-phenotypes. In 4,001 patients admitted to the ICU, they identified three AKI sub-phenotypes using routinely measured laboratory measurements and vital signs. These AKI sub-phenotypes differentiated the risk of dialysis and 28-day mortality¹¹.

Conclusion

In conclusion, substantial progress has been made in the identification of AKI subphenotypes. Different lines of evidence suggest that AKI is not one disease but a collection of many different sub-phenotypes with differing risk factors, underlying pathophysiology and clinical outcomes. While some basic treatment strategies for AKI have improved outcomes, such as prevention of nephrotoxic agents, an improved classification of AKI using biologically relevant sub-phenotypes could lead to enhanced prognostication and predictive enrichment of clinical trials to identify much needed therapeutics for both the prevention and treatment of AKI.